HAL Id: hal-02992198
https://hal.archives-ouvertes.fr/hal-02992198
Submitted on 6 Nov 2020HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
system (SW Indian Ocean)
Ruth Fierens, Samuel Toucanne, Laurence Droz, Gwenael Jouet, François
Raisson, Elisabeth Jorissen, Germain Bayon, Jacques Giraudeau, Stephan
Jorry
To cite this version:
Ruth Fierens, Samuel Toucanne, Laurence Droz, Gwenael Jouet, François Raisson, et al.. Quaternary sediment dispersal in the Zambezi turbidite system (SW Indian Ocean). Marine Geology, Elsevier, 2020, 428, pp.106276. �10.1016/j.margeo.2020.106276�. �hal-02992198�
Quaternary sediment dispersal in the Zambezi turbidite system (SW Indian
Ocean)
Article in Marine Geology · June 2020 DOI: 10.1016/j.margeo.2020.106276 CITATIONS 0 READS 113 9 authors, including:
Some of the authors of this publication are also working on these related projects:
Modeling sediment supply of the Congo watershed since the last 155 kaView project
Submarine landslides in the Northern Tyrrhenian Sea and relationship with the contouritic and turbiditic deposits: morphology, stratigraphy, geotechnics and modelling
View project Ruth Fierens
Université de Bretagne Occidentale 8PUBLICATIONS 20CITATIONS
SEE PROFILE
Samuel Toucanne
Institut Français de Recherche pour l'Exploitation de la Mer 103PUBLICATIONS 2,209CITATIONS
SEE PROFILE
L. Droz
French National Centre for Scientific Research 147PUBLICATIONS 2,506CITATIONS
SEE PROFILE
Jouet Gwenael
Institut Français de Recherche pour l'Exploitation de la Mer 107PUBLICATIONS 1,268CITATIONS
SEE PROFILE
All content following this page was uploaded by Stéphan J. Jorry on 29 September 2020.
Quaternary sediment dispersal in the Zambezi turbidite system
(SW Indian Ocean)
Ruth Fierens, Samuel Toucanne, Laurence Droz, Gwenael Jouet,
François Raisson, Elisabeth L. Jorissen, Germain Bayon, Jacques
Giraudeau, Stephan J. Jorry
PII:
S0025-3227(20)30164-X
DOI:
https://doi.org/10.1016/j.margeo.2020.106276
Reference:
MARGO 106276
To appear in:
Marine Geology
Received date:
19 December 2019
Revised date:
9 June 2020
Accepted date:
14 June 2020
Please cite this article as: R. Fierens, S. Toucanne, L. Droz, et al., Quaternary sediment
dispersal in the Zambezi turbidite system (SW Indian Ocean), Marine Geology (2020),
https://doi.org/10.1016/j.margeo.2020.106276
This is a PDF file of an article that has undergone enhancements after acceptance, such
as the addition of a cover page and metadata, and formatting for readability, but it is
not yet the definitive version of record. This version will undergo additional copyediting,
typesetting and review before it is published in its final form, but we are providing this
version to give early visibility of the article. Please note that, during the production
process, errors may be discovered which could affect the content, and all legal disclaimers
that apply to the journal pertain.
Quaternary sediment dispersal in the Zambezi turbidite system (SW Indian
Ocean)
Ruth Fierensa*, Samuel Toucanneb, Laurence Droza, Gwenael Jouetb, François Raissonc, Elisabeth L.
Jorissena,d, Germain Bayonb, Jacques Giraudeaue & Stephan J. Jorryb
a IUEM-UBO-CNRS, UMR6538, La boratoire Géosciences Océan, 29280 Pl ouzané, France.
b IFREMER Centre Bretagne, Unité de Recherche Géosciences Ma rines, 29280 Pl ouzané, France.
c
Tota l, Centre Scientifique et Technique Jean Féger, 64000 Pa u, France
d Pres ent a ddress: Utrecht University, Department of Earth Sciences, Budapestlaan 17, 3584 CD Utrecht, The Netherlands
e EPOC, UMR 5805, Uni versité de Bordeaux, CNRS, 33615 Pessac CEDEX, Fra nce
*Corres ponding author: Ruth.Fierens@univ-brest.fr
Abstract
This study investigates the Late Quaternary sediment distribution of the Zambezi turbidite system (Mozambique Channel, Western Indian Ocean) from a set of piston cores that characterizes the sedimentation in the Intermediate Basin and in the proximal and distal parts of the Zambezi Fan. Sedimentological and geochemical analyses permit to define variations in sediment composition, sediment accumulation rates and timing of turbiditic deposits over the past 720 kyr. Our study re ve al s low sediment inputs and rare turbiditic deposits in the Zambezi turbidite system, and the de e p (>2,500 m) Mozambique Channel in general, over the studied time interval. The reconstruction of the terrigenous flux in the upper part of the system suggests monsoon-related precipitation changes as the main forcing for riverine inputs variability in the Zambezi system. However, the occurrence of turbi diti c deposits in the cores suggests that there is no genetic link between the ir triggering and evolving climate
and sea-level conditions, thus emphasizing transformation of failed (slide-generated) sediment from the continental slope as the main initiation process for turbidity currents in the Zambezi system. Finally, our data highlight regional-scale changes in sedimentary facies through time, interpreted as successive ‘on-off’ switches in the activity of the distal Zambezi Fan, and by extension, regional-scale depocenter shifts. The last one likely occurred at 350±42 kyr, and is concomitant with a significant increase in terri genous inputs into the proximal Intermediate Basin. It is speculated that this depocenter shift is related to a major southward migration of the Zambezi delta.
Keywords: Zambezi turbidite system, Mozambique Channel, climate changes, monsoon, Quaternary,
depocenter shift
1. Introduction
Turbidite systems are currently a major focus of industry and academic research because they are targets for oil and gas exploration (Shanmugam and Moiola, 1988; Weimer and Link, 1991; Clark and Pickering, 1996; Stow and Mayall, 2000; Pettingill and Weimer, 2002; Mayall et al., 2006; Howell e t al . 2014; Fonnesu et al, 2020) and can hold a potential record of past terrestrial environmental changes (Hesse and Khodabakhsh, 1998; Bouma, 2001; Piper et al., 2007; Clift et al., 2008; Ducassou et al., 2009; Toucanne et al., 2008, 2012; Covault et al., 2010; Jorry et al., 2011; Bonneau et al., 2014). The main forcing factors that influence sedimentary distribution and flux in turbidite syste ms (te ctonics, climate, sea-level, hydrodynamics, deep currents, etc.) have been identified for a l ong ti me , though their exact role and interplay remains debated (see Hinderer, 2012 and Romans et al., 2016 for thorough reviews). Initially, the architecture and evolution of turbidite systems and, more ge ne rall y, the sedimentation on the margins were globally envisioned to be mainly controlled by glacio-eustati c sea-level variations (Vail et al., 1977; Shanmugam and Moiola, 1982; Shanmugam et al., 1985).
According to the principle of sequence stratigraphy, the turbidite system develops mainly duri ng se a-level falls and lowstands, when the rivers, as the main source of these depositional systems, are abl e to reach the shelf edge and deliver sediments directly to the deep basin (e.g., Vail et al., 1977; Posamentier and Kolla, 2003). However, an increasing number of studies provide evidence s that in specific climatic conditions (at both high- and low-latitude) the growth of turbidite systems is also governed by climate that modulates the fluvial discharge and sediment flux from rivers (e .g., Covault and Graham, 2010 for a review). At high latitudes, maximal sediment flux occurs during deglaci al se a-level rise when sediment discharge increases in response to the decay of ice sheets ( Kolla and Permuttler, 1993; Skene and Piper, 2003; Zaragosi et al., 2006; Tripsanas et al., 2008; Toucanne et al., 2008, 2012). At lower latitudes, the intensification of the summer monsoon during interglacial intervals results in higher water and sediment discharges to the ocean, which cause higher turbidity current activity in the deep sea despite sea-level highstand conditions (Ducassou et al, 2009; Pri ns e t al., 2000; Bourget et al., 2010; Liu et al., 2016; Picot et al., 2019). Thus, climate-related changes in the seaward sediment flux strongly modulate the sea-level forcing acting on the deep-sea turbidite systems.
Despite being one of the largest turbidite systems in the world, the feeding and evolution of the Zambezi turbidite system still remains unknown. The study of the Zambezi turbidite system started during the eighties by Kolla et al. (1980a, b), who did a regional study on the sedi ment distribution patterns and physiography. This was followed by the research of Droz and Mougenot (1987) that focused on the various evolution stages of the Mozambique Fan since the Oligocene. More re ce ntl y, studies focused on the geomorphological character of the Zambezi Channel (e.g., Wiles et al. 2017a, b; Breitzke et al, 2017; Castelino et al., 2017; Flemming and Kudrass, 2018; Fierens et al., 2019), but despite increasing research efforts, the knowledge of the sedimentation in the Mozambique Channel is still very patchy. The only recent sedimentary data acquired in the Mozambique Channel are
predominantly concentrated on the Mozambican platform, by focusing exclusively on the last gl aci al -interglacial transition (Just et al., 2014; Schulz et al., 2011; van der Lubbe et al., 2014). In this paper, we fill the gap between geomorphological studies (e.g., Kolla et al, 1980b; Fierens et al., 2019) and l ong-term geological reconstructions (Walford et al., 2005) by addressing which forcing factors contribute to the temporal distribution and depositional patterns of terrigenous sediments of the Zambezi turbi dite system over the last 720,000 years. Specifically, we have analyzed geochemical, sedimentological and stratigraphic characteristics of 11 piston cores located at five strategic localities in the Mozambique Channel. Our results demonstrate the complexity of the Zambezi turbidite system, with a major distal -to-proximal shift in depocenter at ca. 350 ka.
2. Regional setting
2.1. The Mozambique Channel and the Zambezi turbidite / depositional system
The Mozambique Channel (Fig. 1) is located in the western Indian Ocean between Madagascar and the southeast African coast. This north-south trending basin deepens to the south from 2700 m (17°S) to 5000 m (30°S) and has a width that ranges from 430 km at 17°S to 1300 km at 28°S (Breitzke et al., 2017) (Fig. 1A). The Mozambique Channel formed during the break-up of the Gondwana supercontinent in Early Jurassic to Early Cretaceous, when Africa and Antarctica continental blocks separated (Thompson et al., 2019). From Late Jurassic to Early Cretaceous, Madagascar drifted southward relative to Africa along the proto-Davie Fracture Zone to its present position (Leinweber and Jokat, 2012; Mahanjane, 2014). This breakup led to the formation of the N-S trending Davie Ridge (south of 9°S) and the Mozambique and Madagascar Ridges that are located in the southern extension of the Mozambican and Madagascan margins, respectively. The central part of the Mozambique Channel is occupied by se ve ral isolated coral reef islands, namely the Iles Eparses (Jorry et al., 2016) (Fig. 1B), which are inherited from
the long-term development of carbonate platforms on top of volcanic edifices, from Oligocene to Earl y Miocene (Courgeon et al., 2016).
The Zambezi turbidite system covers a major part of the Mozambique Channel. The principal source of sediment to this sedimentary system is the Zambezi River (Fig. 1), which is the fourth largest river in Africa and the largest flowing into the Western Indian Ocean (Walford et al., 2005). It has a catchment area of 1.3x106 km2, a mean elevation of ~1 km, a total suspended sediment of 48 Mt.yr-1 and drains the
southern part of the East African Rift System as well as the South African plateau (Thomas and Shaw, 1988; Summerfield and Hulton, 1994; Milliman and Farnsworth, 2011) (Fig. 1). Walford et al. (2005) defined a long-term (120 Ma) record of sediment supply of the Zambezi River towards the Mozambique margin using a two-dimensional seismic reflection survey. These authors propose a possible relationship between increasing sediment flux during Pliocene and an enlargement of the Zambezi watershed area. It is important to note, however, that the Zambezi River drainage reorganization during the Plio -Pleistocene is very complex (Nugent, 1990; Thomas and Shaw, 1991; Moore and Larkin, 2001; Cotte ri l l , 2006; Moore and Cotterill, 2010; Moore et al., 2012) and is still very much debated. In addition, the sediment flux in Southeast Africa has also been linked to stages of hinterland uplift that are deemed responsible for large amounts of mass wasting on the slope and shelf ( Green, 2011a, 2011b; Wiles et al., 2013; Hicks and Green, 2017; Micallef et al., 2019).
The Zambezi turbidite system consists of two different depositional systems: the Zambe zi Fan (Fi g. 2) and the Intermediate Basin (Figs. 1 and 3) (Fierens et al., 2019). The Zambezi Fan is characterized by the Zambezi Valley and its tributaries (e.g. Tsiribihina Valley), and by its distal distributaries and depositional area (i.e., the fan sensu stricto). The Zambezi Valley is wide, mostly rectilinear and originates in the north from the Mozambique margin (Wiles et al., 2017b). At present, the submarine valley is disconnected from the Mozambican shelf and only appears 280 km eastward from the Zambezi River outlet (Schulz e t al., 2011; Wiles et al., 2017a; Fierens et al., 2019). The upstream part has a NW-SE orientation, which i s
transverse to the Mozambican margin. Further to the east, the Davie Ridge forms a morphological barrier making the valley change to an N-S trend (Schulz et al., 2011). The Zambezi Valley me rges wi th the Tsiribihina Valley, which originates from the Madagascar margin, at around 22°S. The Zambezi Valley is included in the inner fan by Kolla et al. (1980b) (zone I in Fig. 1B). At the southern limit of the Zambezi Valley (~25.5 °S), an ancient channel-levee system is identified (Fig. 2B, C; Fierens et al., 2019). Downstream of this the submarine valley ends in the Mozambique Basin where lobe deposits can accumulate (line 4, Fig. 2D) (Kolla et al., 1980b; Fierens et al., 2019). Kolla et al. (1980b) subdivi de d thi s distal deposition area into a middle and outer fan (zone II and III in Fig. 1B). The middle fan is defined as the area where the valley expands into several distribution channels where sand deposits predominate . The outer fan is located in the most distal part of the Mozambique Basin where distributary channels are absent and where finer grained sediments are recognized (Kolla et al., 1980b). In the study area, Kolla e t al. (1980a) and Breitzke et al. (2017) recognized large-scale seafloor bedforms, which are related to strong bottom-current influence, as also recently reported by Miramontes et al. (2019). Additionally, the Mozambican continental slope, the Zambezi Valley and the Iles Eparses encompasse s a ponde d se mi -confined basin which is referred to as the Intermediate Basin (Fierens et al., 2019).
2.2. Oceanography and climate
A complex and dynamic oceanic circulation pattern characterizes the Mozambique Channel, consi sti ng of large surface gyres and deep oceanic currents (Fig. 1A). Along the Mozambique margin, a train of deep-reaching anticyclonic eddies propagates southward. These eddies appear to be over 300 km wi de and cause a net poleward transport of about 15x106 m3/sec in the upper 1500 m (de Ruijter et al., 2002;
Schouten et al., 2003). The eastern side of the channel is characterized by temporary cyclonic eddies that move from the southern tip of Madagascar in west-southwesterly direction (Quartly and Szokosz, 2004). The deep oceanic circulation is composed of the Antarctic Bottom Water (AABW) and North
Atlantic Deep Water (NADW) and the Antarctic Intermediate Water (AAIW). The AABW (below 4000 m) and NADW (2000 m - 3500 m) flow in a northward direction along the western margin of the Mozambique Channel. At 25°S, the bathymetry becomes shallower and the AABW gets deflected towards the east and south. Similarly, the lower portion of the NADW gets deflected towards the south due to decreasing topographic depths at 20°S. The upper part of the NADW forms together with the AAIW (800 m – 1500 m) the Mozambique Undercurrent (MUC), which has a mean northward speed of 4.6 cm/s (1500 m) and 4.5 cm/s (2500 m) (Ridderinkhof and de Ruijter, 2003). This undercurrent continues flowing northwards, crosses the Davie Ridge and flows into the Somalie Basi n (de Rui jter e t al., 2002; Donohue and Toole, 2003; Ridderinkhof and de Ruijter, 2003; Van Aken et al., 2004; Breitzke et al., 2017).
The climate of the East Africa region is governed by the movement of the Intertropical Convergence Zone (ITCZ). On seasonal timescales, the ITCZ migrates meridionally causing an equatorial zone of humid climate with a high degree of inter-annual rainfall variability. The Zambezi catchment (and the watersheds of the Madagascar rivers) gets maximum precipitation during austral summer (De ce mber– February) when the ITCZ boundary migrates to the northern half of the Zambezi Catchment (e.g., Ziegler et al., 2013).
The available terrestrial paleoclimate records from the African continent are often either relatively short (e.g., Talbot and Johannessen, 1992; Beuning et al., 1997, Thevenon et al., 2002; Johnson et al. 2002; Kristen et al., 2010) or long but fragmented (e.g., Barker et al., 2002; Trauth et al., 2005). These have been recently complemented by drill cores taken from long-lived lakes and paleolakes, whi ch provi de continuous records of terrestrial climate (Scholz et al., 2007; Lyons et al., 2015; Johnson et al., 2016). However, the availability of such high-resolution continental paleoclimate archives is l i mi ted and the y are typically geographically dispersed (deMenocal, 2004). Hence, paleoclimate studies are commonly based on a single terrestrial or marine sediment core (e.g., Schefuß et al., 2005; Garcin et al., 2006;
Brown et al., 2007; Lyons et al., 2015; Johnson et al., 2016; Castañeda et al., 2016) . Consequently, the climate history of tropical and subtropical southeast Africa and the mechanisms that control rainfall and temperature are still highly debated.
In general, Eastern African climate is considered to have developed from warmer, wetter condi ti ons i n the Late Miocene and Early Pliocene (Cerling et al., 1997; Woldegabriel et al., 2001) towards progressively cooler and drier climate over the past few million years (Bonnefille, 1983; Ce rl i ng, 1992; Potts, 1996; deMenocal, 2004; Trauth et al., 2005). This long term change towards more arid conditi ons is punctuated with step-like shifts in periodicity and amplitude of glacial-interglacial changes. Throughout the Late Neogene the African monsoon climate varied at periodicities of 23–19 kyr (orbi tal precession), which was superimposed to high-latitude glacial cycles of 41 ka after ca. 1.8–1.6 Ma, and of 100 ka after the Mid-Pleistocene Transition (MPT; ca. 1.2–0.8 Ma) (deMenocal, 2004).
This general assumption is not supported by the core data from Lake Malawi (10° to 14°S, Fig. 1A), the southernmost lake in the West African Rift Valley. This lake could persists through severe and prolonged arid periods as it is one of the deepest and oldest lakes in the world, making possible a conti nuous 1.3-Myr continental record (Fig. 4; Lyons et al., 2015; Johnson et al., 2016). Based on temperature and aridity records from Lake Malawi, Johnson et al. (2016) confirmed the 41-kyr climate periodicity prior to the Mid-Pleistocene transition. After the MPT, the environment became progressively more humid with strong 100-kyr cycles consisting of warm, wet interglacial periods and cool, dry glacial intervals ( Fi g. 4A; Johnson et al., 2016). This is supported by additional paleoclimate proxies that suggest lower lake levels with frequent water level changes before the MPT followed by overfilled lake conditions (reflecting wetter climate) that were interrupted by intervals of extreme drought until ca. 100 kyr (Fig. 4B; Lyons e t al., 2015). By correlating core results of Lake Malawi (11°S) with seismic data, Scholz et al. (2007) specified that these extreme drought periods occurred during insolation minima. This is consistent wi th previous research on the Pretoria Saltpan in South Africa (Fig. 4C; Partridge et al., 1997).
These results are, in turn, partly contradicted by vegetation and hydroclimate reconstructions of the Limpopo catchment area (20°-24°S). Here, a long-term aridification was found between 1 and 0.6 Ma, followed by more-humid conditions with no evidence of the megadroughts (Fig. 4A; Castañeda et al., 2016; Caley et al., 2018). In addition to that, precipitation proxies on a core from Lake Tanganyika (~6°S) indicate numerous abrupt changes in hydrology throughout the past 60 kyr (Tierney et al., 2008).
3. Materials and methods
3.1. Core locations
This study is based on 11 piston cores recovered from the Mozambique Channel between 18°S and 27°S latitude during the MOZ1 (Olu, 2014), MOZ2 (Robin and Droz, 2014) and PAMELA-MOZ4 (Jouet and Deville, 2015) oceanographic cruises (see Fig. 1B for piston cores location). The location of the cores was determined first to gain the best sedimentological insight into the main physiographic margin (e.g. slope, abyssal plain) and turbidite entities (i.e., fan, Inte rme di ate Basi n) of the Mozambique Channel, from the Mozambique and Madagascar slopes to the North to the deepest part of the Zambezi sedimentary system to the South. Secondly, they have been retrieved from (or along) the main sedimentary and geomorphological features (levees, lobes, channels, etc.) re cognized on high-resolution multibeam and CHIRP data (Figs. 2 and 3). The sediment cores range from 4.2 to 15.3 meters in length and are located in water depths between ca. 2,500 and 4,400 m. Details about the cores analyzed in this study are displayed in Table 1.
3.2. Sedimentology
Sedimentary facies (Table 2) were described according to sediment texture and composition. For this we performed visual description, high-resolution digital photography and X-ray radiographies (Geotek-XCT system) over the entire length of the sediment cores. Additionally, grain-size analysis (n = 521) were
done on selected depth intervals. Bulk samples were measured using a Mastersizer 3000 laser diffraction particle size analyser coupled to a Hydro LV wet dispersant unit. Finally, the CIE L*a*b* col or indices were determined at 1 cm resolution by a Konica Minolta CM 2600d coupled to a GEOTEK Mul ti -Sensor Core Logger (MSCL). Acquisitions were carried out with a measurement area of 8 mm and using a daylight illuminant (D65) and 10° viewing angle (specular component excluded), directly from the spl i t surface of the cores. In this study, we focus on the a* parameter (red-green scale) of the spectrophotometric results, which is a good indicator for the color differences occurring along the piston cores.
3.3. Geochemistry
In order to constrain the mechanism of color changes in our sediment cores, we performed detailed geochemical analyses. Semi-quantitative determinations of geochemical profiles for major elements were acquired by an X-ray fluorescence (XRF) Avaatech core scanner. Measurements were done on each piston core and performed every centimeter with a counting time of 10 sec. The chemical elements ( Al, Si, Ca, Ti, Fe) were detected by using two acceleration voltages (10 kV and 30 kV, respectively). The Al/Si and Fe/Ti ratios reflect variations in sediment delivery, while the ratio of Ca/Fe is indicative of bi oge nic primary productivity and, by extension, of past climate variability (see Richter et al., 2006; Rothwe l l e t al., 2006; Govin et al., 2012 for thorough reviews). Additionally, quantitative determination of total CaO concentrations were obtained by a Bruker S8 Tiger CRF spectrometer in 10 selected sediment sampl es from core MOZ4-CSF20 (Zambezi slope). Both cores retrieved from the Zambezi Valley (MOZ1-KSF30 and MOZ2-KS11) were analyzed for their total organic carbon (TOC) content using the Shimadzu Total organic carbonate (TOC-L CSH) analyser with a NDIR (Non-Dispersive Infra-Red) sensor. Sampl es we re taken with a step size of 20 cm, dried (50°C), crushed and decarbonized using HCl (4%). TOC content was calculated by subtracting the Inorganic Carbon (IC) content from the Total Carbon (TC) content. TC is
determined on samples that underwent combustion catalytic oxidation at 680°C and IC is detected on samples that were acidificated by phosphoric acid.
Finally, 25 samples were analyzed for their Neodymium isotopic composition (Nd) in order to deci phe r potential changes in the Zambezi River drainage reorganization through time ( e.g., Thomas and Shaw, 1991; Cotterill, 2006; Moore et al., 2012). The neodymium isotopic composition (εNd) of terrigenous sediment is a powerful tracer for geographical provenance because the εNd signature of detrital sediment is retained during continental weathering and subsequent transport (Goldstein and Jacobse n, 1988; Bayon et al., 2015). Samples were taken at selected depth intervals on 7 cores (MOZ1-KS26, MOZ4-CS22, MOZ2-KS05, MOZ1-KSF30, MOZ2-KS11, MOZ2-KSF15 and MOZ2-KSF13). For full methodology on the sequential leaching procedure, alkaline fusion and isolation of Nd fractions we refer to Bayon et al. (2002), Bayon et al. (2009) and Bayon et al. (2012). Isotopic measurements were performed at the Pôle Spectrometrie Ocean, Brest (France), using a Thermo Scientific Ne ptune mul ti -collector ICPMS. Mass bias corrections on Nd were made with the exponential law, using (146Nd/144Nd) =
0.7219. Nd isotopic compositions were determined using sample-standard bracketing, by anal yzing an in-house SPEX Nd standard solution (SPEX) every two samples. Mass-bias corrected values for
143Nd/144Nd were normalized to a SPEX Nd value of (143Nd/144Nd) = 0.511709. Replicate anal yses of the
JNdi-1 standard solution during the course of this study gave (143Nd/144Nd) = 0.512116 ± 0.000011 (2SD,
n = 10), in full agreement with reference value (0.512115) from Tanaka et al. (2000), and correspondi ng to an external reproducibility of ~ ±0.21 ε-units taken as the estimated uncertainty on our measurements. In this study, both measured 143Nd/144Nd ratios and literature data are reporte d i n Nd
notation [(143Nd/144Nd)
sample/(143Nd/144Nd)CHUR-1]x 104, using (143Nd/144Nd)CHUR = 0.512630 (Bouvier e t al .,
2008).
3.4. Stratigraphy
Accurate age models have been constructed for all cores based on stable isotopes (C/O), radiocarbon acceleration mass spectrometry (AMS) 14C dating and nannofossil age determinations.
Foraminifera used for stable isotopes and 14C datings were found along the hemipelagic layers with a
sample step of 5 to 20 cm. These hemipelagic layers are not contaminated by reworked material and represent intervals of continuous sedimentation. The samples were dried and wet-sieved at 63 μm. For oxygen isotope analysis about 100 μg of the benthic foraminifera species Cibicides wuellerstorfi were picked with a size fraction between 250 and 425 µm. The isotopic composition was analyzed at the Leibniz Laboratory for Radiometric Dating and Stable Isotope Research (Kiel University, Germany) usi ng a Kiel IV carbonate preparation device connected to a MAT 253 mass spectrometer from ThermoScientific. During preparation the carbonates were reacted with 100% phosphoric acid (H3PO4) under vacuum at 75°C, and the evolved carbon dioxide was analyzed eight times for each individual sample. All values are reported in the Vienna Pee Dee Bee notation (VPDB) relative to NBS19. Pre ci si on of all different laboratory internal and international standards (NBS19 and IAEA-603) is <±0.05 ‰ for δ13C and <±0.09 ‰ for δ18O values. For core MOZ4-CS25 the δ18O measurements are supported by 3
nannofossil ages. These analyses are based on biostratigraphic and age markers from Leg ODP 175 (Shipboard Scientific Party, 1998). Core stratigraphy has been completed for upper core sections by Radiocarbon dating conducted on 27 selected samples (Table 3). Analyses were performed at the BETA Analytics laboratory (USA) on shells of planktic foraminifera ( monospecific assemblage of G. ruber or bulk planktic foraminifera) with a fraction between 250 and 315 μm. Radiocarbon ages were first corrected for reservoir age and then calibrated to calendar age using the IntCal13 calibration curve (Reimer et al., 2013). Reservoir correction was inferred from radiocarbon measurements in prebomb known-age shells and coral from the tropical SW Indian Ocean (Southon et al., 2002) and calculated according to the uncalibration-convolution process of Soulet (2015), giving a current regional re se rvoir
age of 426 ± 76 14C years. Due to the lack of knowledge about changes of this regional reservoir age
through time, this value was used to correct all 14C ages of this study. Finally, the stratigraphic
framework allows the calculation of linear sedimentation rates and mass accumulation rates.
3.5. Mass accumulation rates
Terrigenous supplies were quantified for core MOZ4-CSF20 by the calculation of the mass accumulati on rates (MAR or terrigenous flux). This core was selected because of its high sedimentation rate s and i ts ‘proximal’ position relative to the Zambezi terrigenous sediment source (see Section 5.2). For this purpose, the coring artefacts were first corrected based on the CINEMA software developed at IFREMER (Bourillet et al., 2007). The MAR, in g cm-2 kyr-1, were subsequently calculated for glacial and interglacial
periods according to the following formula:
MAR = LSR * DBD * (1 carbonate content),
with LSR for Linear Sedimentation Rate (cm kyr-1) and DBD for Dry Bulk Density (g cm-3). The latter has
been calculated by the formula DBD = 2.65 * (1.024 Dwet)/(1.024 2.65), where 2.65 equals the mean
grain density (in g cm-3) and 1.024 is the interstitial water density in g cm-3 (Auffret et al., 2002). The we t
bulk densities (Dwet) were derived from gamma-ray attenuation density measurements.
4. Chronostratigraphic framework
Reliable age models for 10 cores (on 9 coring sites; see Table 1 and the associated caption) were constructed by synchronizing the C. wuellerstorfi 18O isotopes records with the global benthic
foraminiferal LR04 stack of Lisiecki and Raymo (2005) (Fig. 5). Results obtained from cores MOZ1-KS26 and MOZ4-CS22 (twin cores, i.e., taken at the same location; see Table 1) are combined to establish a composite age model, with the upper part of the record corresponding to MOZ1-KS26. In the text and
figures that follow, we thus use the name MOZ1KS26 to denote this composite record. For tuni ng, ti e -points were limited to clearly identifiable marine isotope stage (MIS) boundaries (Lisiecki & Raymo, 2005; Railsback et al., 2015), including Terminations (i.e., glacial-interglacial transitions) ( Tabl e 4). The chronology of the uppermost portion of the cores is supported by 27 radiocarbon ages (Table 3). For core MOZ4-CS25, additional nannofossil ages were obtained at 81, 167 et 420 cmbsf indicating MIS 12-15 (ca. 460 – 600 ka). This result is based on the acme (i.e., intervals of single species dominance of the flora) of G. caribbeanica, the presence of P. lacunose and the absence of R. asanoi and H. sellii. The age model of core MOZ4-CSF20 encompasses the last 162 ka (i.e., core bottom corresponding to mid-MIS 6). Meanwhile, the age model of core MOZ2-KS05 revealed that this core extends back to ca. 720 ka (MIS 18). The other seven cores register a maximum age that lie in between these two ages (Table 4 and Fi g. 5).
The glacial/interglacial amplitude of our δ18O data (ca. 2 ‰ over the last 4 climate cycles) closely
matches with the standard LR04 stack (Fig. 5). However, the stable oxygen isotope values generally vary between 2.5 and 4.5 ‰, making them consistently 0.5 ‰ lower than the LR04 record. This offset is probably related to the different water depths and local signature of our cores (3553±1052 m depth) with respect of the signature of the cores used to produce the LR04 stack (3127±1992m depth). Additionally, sediment archives recovered from the distal part (MOZ2-KSF15, MOZ2-KSF13, MOZ2-KS11 and MOZ1-KSF30) demonstrate slighter higher values per mil VPDB than cores taken in the northern part of the study area (MOZ4-CSF20, MOZ1-KS26, MOZ2-KS05 and MOZ1-KSF24). In accordance with the findings of McCave et al. (2005), this lateral variation of the benthic δ18O signature likely results from the
different deep water masses present at these sites (AABW and NADW, respectively).
5. Results
5.1. Lithofacies
Six main sediment facies have been inferred from visual description, X-ray radiographs and grai n-si ze analysis (Table 2). Core MOZ4-CS21, sampled on the Zambezi slope, is predominantly composed of homogeneous dark grey silty-to sandy-clay (Fig. 6). These sediments are characterized by poor sorti ng, frequent inverse grading and a general bimodal grain-size distribution. Consequently, these muds solely found in core MOZ4-CS21 can be interpreted, according to Stow and Lovell (1979) and Stow and Piper (1984), as contouritic deposits, i.e., sediments deposited or substantially reworked by the persistent action (e.g., selective deposition, winnowing, erosion) of bottom currents ( Rebesco et al., 2014) . Thi s i s supported by radiocarbon ages that suggest longer settling time, which excludes the possi bi li ty for an instantaneous gravity flow deposit (Fig. 6). The other sediment cores from the Mozambique Channel consist predominantly of hemipelagic deposits (Fig. 7), which are widely distributed throughout the entire study area. This facies shows significant color variability between the yellowish beige versus dark grey types, and consists of bioturbated mud (minor grain size mode = ~4 μm) with abundant foraminifera (major grain size mode of ~450 μm). The hemipelagic sediment at proximal si te s (MOZ4-CSF20, MOZ4-CS21, , MOZ1-KS26, MOZ2-KS05, MOZ1-KSF24 and MOZ2-KSF13) consists predominantly of olive to dark grey, foraminifera-rich silty-clay (Fig. 8) which is characterized by a total organic conte nt (TOC) of up to 0.5 % (Fig. 9) and a moderate to high sedimentation rate (up to 9 cm/kyr at MOZ4-CSF20; Fig. 10). As a prominent example, MOZ2-KSF15 demonstrates that this hemipelagic grey facies (between 7 m and 10 m in particular in Fig. 11E, F) is enriched in iron and aluminum (relative to calcium; Fi g. 11C, D) and has higher sedimentation rates (i.e., two-fold increase relative to the biogenic-dominated faci es; Fig. 11D). Yellowish beige hemipelagic deposits dominate in the distal Zambezi depositional system (e.g., cores MOZ1-KSF30, MOZ2-KS11, MOZ2-KSF15, MOZ4-CS25) (Fig. 8). It represents a 3 to 7 m thick sediment cover at the studied sites (Fig. 8), including MOZ1-KSF30 and MOZ2-KS11 that are retrieved on top of a levee of an ancient (i.e., non-active) channel-levee system located at the e dge of the pre se nt
Zambezi Channel (Fig. 2). This sediment is characterized by low TOC (~0.1 %) and very low sedimentation rates (<2 cm/kyr) (Fig. 9).
Hemipelagic deposits are interbedded by mainly fine-grained or massive sandy turbidites that can be easily identified by their coarseness (except at site MOZ2-KS05, see below), normal grading and disti nct erosive bases (e.g., Stow & Piper, 1984). Fine-grained turbidites (Fig. 12) have a relative widespread occurrence in the turbidite system as they are observed on the Mozambique slope (MOZ4-CSF20) and Madagascar slope (MOZ1-KSF24 and MOZ2-KSF13), in the Intermediate Basin (MOZ1-KS26, MOZ4-CS22 and MOZ2-KS05) and on the lower Zambezi Valley flanks (MOZ1-KSF30 and MOZ2-KS11). The recurre nt, fine-grained turbidite successions found at proximal sites (e.g., cores MOZ4-CSF20 and MOZ1-KSF24) and the scarcity of turbidites in the distal Zambezi depositional system (e.g., core MOZ2-KSF15) indicates a proximal-distal trend (Fig. 8). They are characterized by a median (D50) grain-size of 4 to 122 μm (47122 μm for the basal part) and a ninety percentile (D90) ranging from 22 to 642 μm (109642 μm for the basal part). These fine-grained turbidites are interbedded within the olive to dark grey hemipelagic deposits. It is worth noting that in core MOZ2-KS05 (Intermediate Basin) turbidites are remarkably finer-grained compare to elsewhere, so that the turbidite beds are embedded by coarser (foraminifera-rich) hemipelagic facies (Fig. 13). Also within the Intermediate Basin, fine-grained turbidites can occur in a stacked configuration showing ripples with different current directions (Fig. 12). This may suggest an effect of rebound and ponding processes (Remacha and Fernández, 2003; Ti nte rri and Muzzi Magalhaes, 2011). Massive sandy turbidites (Fig. 14) are mainly observed in the deeper parts of the system, i.e., in the distal fan at sites MOZ2-KS11 and MOZ4-CS25. This may be related to an overall bypass of turbidity currents in proximal fan locations. They are interbedded in the yellowish hemipelagic facies (e.g., Fig 14). They are distinctly coarser grained with a median (D50) grain-size of 54– 239 μm and a ninety percentile (D90) ranging 124–689 μm. Locally, two particular turbidite facies are recognized. A 14-cm thick, organic-rich turbidite (Fig. 15) is solely observed in core MOZ2-KS11.
Secondly, carbonate turbidites (Fig. 16) are observed mainly in core MOZ1-KSF26 and MOZ4-CS22, which are located in the vicinity of the isolated carbonate platforms forming the Iles Eparses. These carbonate turbidites are originated from the shedding of Bassas da India atoll top, as previously described by Counts et al. (2018). Occasionally, thin carbonate turbiditic beds are found on the core s of Madagascar margin (MOZ1-KSF24 and MOZ2-KSF13).
5.2. Sedimentation rates, fluxes and turbiditic deposit frequencies
Based on the age models, sedimentary accumulation rates (SAR) have been determined for each gl aci al and interglacial period (according to the MIS boundaries of Lisiecki & Raymo, 2005) (Fig. 10).
Overall, for all cores, relatively low sedimentation rates are found. SAR varies between 0.13 cm/ka (MOZ2-KS05) and 9.64 cm/ka (MOZ4-CSF20) and no considerable variations are observed between glacial and interglacial periods. The data also indicate a general decrease of mean sedimentation rate with increasing distance from the source areas. The highest age-depth profile gradient and mean sedimentation rate (6.1 cm/ka) is found most proximal at the Mozambique margin (MOZ4-CSF20) ove r the past ca. 160 kyr (Fig. 10). Further to the south, the proximal Intermediate Basin (MOZ1-KS26) records a mean SAR close to 3.0 cm/ka during MIS 13 to MIS 1. More distal ly in the Intermediate Basin (MOZ2-KS05) the mean SAR value is similar over the last 243 ka (MIS 7 to MIS 1). Before (between MIS 18 to MIS 8), a very low mean sedimentation rate of 0.3 cm/ka is found. The cores from the distal Mozambique Basin (MOZ1-KSF30, MOZ2-KS11 and MOZ2-KSF15) record average sedimentation rates close to 2.0 cm/ka (and distinct lower profile gradients) over the past 720 ka. It is interesting to note that increased sedimentation rates (to 3.6 cm/ka) from 715 to 910 cmbsf in the Zambezi lower fan site is coeval with the deposition of a Fe- and Al-rich, dark grey hemipelagic facies (Fig. 11C to 11E). Finally, the
cores retrieved from the Madagascar margin (MOZ1-KSF24 and MOZ2-KSF13) show mean sedimentation rates of 2.3 cm/ka during the past ca. 350,000 yr.
The ‘proximal’ location of core MOZ4-CSF20, added to the high sedimentation rates (up to ~10 cm/ka) and the recognition of the main stratigraphic 18O features (e.g., MIS 5a to MIS 5e) over the last ca. 162
ka led to quantify sediment flux (i.e., mass accumulation rates, MAR) at this site in order to expl ore the possible relationship existing between climate changes and terrigenous inputs over the last climate cycle (Fig. 17). The MAR shows a significant variability over the last climatic cycle. The results reveal that MAR ranges from 2-3 g.cm².kyr-1 during glacial intervals (MIS 2, MIS 6) to 5-7 g.cm².kyr-1 during interglacials
(MIS 1, MIS 5), with MAR peaks corresponding to the interglacial warm intervals (i.e., MIS 5a, c and e; Fig. 17).
Finally, the sediment inputs in the turbidite plains have been examined through the turbidite occurrences and the calculation of the turbidite frequencies (Fig. 18). For all individual cores, the duration of each isotope stage (kyr) is divided by the amount of turbidites per isotope stage, thus giving turbidite recurrence times per isotope stage as presented through white italic numbers in figure 18. For the ease of comparison, additional Gaussian shaped frequency plots are constructed . Their height increases when the turbidite frequency increases. Overall, turbidites found within the hemipelagic deposits occur irrespective of glacial or interglacial periods, i.e., intervals of high and low sea-level respectively (Fig. 18).
In the Zambezi depositional system, both proximal and distal sites indicate a low turbidites occurre nce from MIS 13 to MIS 9 (turbidite recurrence times of ca. 40 kyr). During the last ca. 300 ka, the turbidite activity increases at proximal sites (Zambezi slope and Zambezi Intermediate Basin sites; turbidite recurrence times of ca. 23 kyr) but decreases substantially (i.e., only turbidites in core MOZ2-KS11) in the distal part of the system (Zambezi Valley and Zambezi Lower Fan). During MIS 6 and 8, turbi di te s i n
the Intermediate Basin (i.e., MOZ1-KS26; Fig. 12) are mainly present in a stacked configuration (i.e., one upon another) with ripples characterized by multiple directions, which may be the result of local rebounding and ponding processes (Remacha and Fernández, 2003; Tinterri and Muzzi Magalhaes, 2011).
On the other side of the Mozambique Channel (cores MOZ1-KSF24, MOZ2-KSF13), fine-grained turbidites are interbedded within the hemipelagic deposits that spans the past ca. 290 kyr (Fig. 18). These cores register similar to slightly higher turbidite activity (average recurrence time of 18 kyr ove r the last 300 kyr) with a maximum activity (turbidite recurrence times of ca. 10 kyr) during MIS 2-4 (MOZ1-KSF24), MIS 5 and 7-8 (MOZ2-KSF13).
5.3. Nd isotope analysis
Nd results are used as a proxy for changes in the Zambezi drainage system (see Methods). The Nd values obtained in the hemipelagic deposits of the studied cores range from -12.2 to -20.6 (n=25) (Tabl e 5) and the dataset show remarkable homogeneity over the last 720,000 years. Indeed, 20 of the 25 measurements give Nd values ranging from -13 to -15 (whatever the core location and the type of hemipelagic deposit, i.e., yellowish beige versus dark grey, see below), and only two samples show Nd values far from this range (-19.1 for MIS 1 and -20.6 for MIS 10 at sites MOZ2-KS05 and MOZ1-KSF30, respectively). The mean Nd value for the whole dataset is -14.4 ± 1.8. This homogeneity suggests that there was no major drainage reorganization over the past 720,000 years.
6. Discussion
6.1. Tracking detrital sediment input in the Mozambique Channel and the Zambezi
turbidite system
Turbidites found along submarine channels and in deep oceanic basins can be good indicators of fl uvi al inputs (Normark and Piper, 1991; Piper and Normark, 2009; Covault and Graham, 2010) and of their variability over time (e.g., Covault et al., 2010; Jorry et al., 2011; Bonneau et al., 2014). The gene ti c l i nk between turbidites and fluvial sources in the Mozambique Channel is emphasized by the number of turbidites that decreases with distance to the margin (see Section 5.1). This is associated to de cre asi ng sedimentation rates to the distal part of the Zambezi depositional system (see Section 5.2). In addi ti on to turbidites, hemipelagic sediment in turbidite successions significantly contributes to this trend. Considering the proximity of cores MOZ4-CSF20, MOZ1-KS26, MOZ1-KSF24 and MOZ2-KSF13 to terrigenous sediment sources (i.e., Zambezi, Tsiribihina or Onilahy rivers), we assume that suspension/transport of fluvial-derived sediment by oceanographic processes (likely includi ng i nte rnal tides, geostrophic currents; Shanmugam, 2013, 2018; Gardner et al., 2017) participates together with turbidity currents to the terrigenous sedimentation on the studied margins. This has be e n thoroughly discussed for the Mozambique margin and the Zambezi River for the past 20 kyr (van der Lubbe et al., 2014), and our data suggest that this process is effective along the Mozambique and Madagascar margins over the last glacial-interglacial cycles (at least since MIS 13 and MIS 10 at sites MOZ1-KS26 and MOZ1-KSF24, respectively; Figs. 8, 18 and 19). The description of dark grey silty- to sandy-clay showing bi-gradational sequences at site MOZ4-CS21 (Fig. 6, Table 2) suggests, in agreement with previous results of Kolla et al. (1980a, b), that the terrigenous sedimentation along the Mozambique margin is impacted by bottom-currents.
Yellowish beige foraminifer ooze (see section 5.1 for their characteristics) dominates in the distal Zambezi depositional system (e.g. cores MOZ1-KSF30, MOZ2-KS11, MOZ2-KSF15, MOZ4-CS25) and is interpreted as a biogenic-dominated facies (Kolla et al., 1980a, b; Stow and Piper, 1984), constituting the background oceanic sedimentation in this region, i.e., Zambezi Fan is not receiving terrigenous sediments (“off” mode). The massive sandy turbidites (e.g., Fig. 14) present in the upper
dominated sequence of core MOZ2-KS11 (Fig. 8), suggest that the Zambezi Channel was fed by rare, sandy turbidity currents during the past ca. 318 kyr, producing rare levee-overbank deposits. Below the biogenic-dominated sequence of core MOZ2-KS11 (prior to ca. 318 kyr) the presence of olive to dark grey silty-clay sediments and organic-rich, fine-grained turbidites demonstrates terrigenous inputs that contributed to the development of this ancient channel-levee system (Fig. 9). The facies transition, from yellow- biogenic-dominated to grey-terrigenous sediment, is observed at regional scal e (i .e., i n core s MOZ1-KSF30, MOZ2-KS11, MOZ2-KSF15, MOZ4-CS25; Figs. 8, 18 and 19) and occurred between 437±5 ka (MOZ2-KSF15) and 318±2 ka (MOZ2-KS11), depending of the depositional environments (i.e., channel edge, levee, lobes; Fig. 18). Fine-grained turbidites, interpreted as overbank deposits, are descri bed on the levee (core MOZ1-KSF30) while thick, massive sandy turbidites are found downstream at the channel outlet (MOZ4-CS25). No silty or sandy laminae are visible in the grey silty-clay facies at core MOZ2-KSF15. However, its characteristics (composition and sedimentation rates; see Section 5.1) coul d indicate, together with the depositional environment, that this facies is made of stacked, base -cut-out mud turbidites (Stow and Piper, 1984). Moreover, this homogeneous grey facies at core MOZ2-KSF15 (ca. 437±5 to 508±7 ka; Fig. 18) is concomitant to the deposits of terrigenous turbidites at the other sites (i.e., cores MOZ1-KSF30 and MOZ4-CS25; Fig. 18). Hence, we could assume that it highlights terrigenous inputs in the deep Mozambique Basin, and by extension, feeding of the distal Zambezi depositional system (i.e., Zambezi Fan in an “on” mode due to the direct feeding of the Zambezi Valley; Fig. 19). Thus, the description of a new succession of yellow- biogenic-dominated (~10–12 m or ca. 510–640 ka; i.e., “off” mode of the Zambezi Fan) and grey-terrigenous facies in the lower part (> 12 m or > 640 ka; i.e., “on” mode of the Zambezi Fan) of core MOZ2-KSF15 (Figs. 11 and 18) suggests a discontinuous feedi ng, with recurrent periods of quiescent conditions (corresponding to the yellowish beige facies; “off” mode) of the distal Zambezi system through time (Fig. 19). This is also supported by the succession of these facies in the lower part of core MOZ4-CS25 (Fig. 18), west of MOZ2-KSF15 (see Fig. 1B for locations).
Based on the sedimentological considerations described above, a detailed reconstruction of the temporal and spatial distribution of terrigenous deposits in the Zambezi system is proposed below in Section 6.2.
Of course, the low turbidite occurrence in our piston cores does not mean that there was nearly no activity in the Zambezi turbidite system since 720 ka. The difficulty for the turbidity currents to produce overbank deposits along the very deep (ca. 700 m; Fierens et al., 2019) Zambezi Valley may be the reason why turbidites are rarely identified. Moreover, the very coarse material present both on the deep submarine valley floors (see DSDP drilling sites 243-244; Simpson et al., 1974) and in the distal depositional part of the Zambezi sedimentary system (Robin and Droz, 2014) evidence s some ‘re cent’ activity. Since this study is based on only 11 cores taken at specific regions of the turbidite systems, it might be anticipated that the turbidite activity may have been significantly higher than recorde d i n the present study.
6.2. Interpretation of detrital facies in a temporal and spatial context: insights into forcing
factors
6.2.1. Proximal deposition environments
The most complete sedimentary records in the Zambezi turbidite system are provided by deposition sites located in the vicinity of terrigenous sediment sources and are therefore the best re posi tories of information on timing of sediment flux from continent to ocean. Hence, we use cores MOZ4-CSF20, MOZ1-KS26, MOZ1-KSF24 and MOZ2-KSF13 to thoroughly evaluate the evolution of the detrital sedimentary inputs in the Mozambique Channel.
These sedimentary records observed in different proximal depositional environments reveal that turbiditic contributions on the margins occurred in both glacial and interglacial conditions (see Se cti on
5.2) (Fig. 18). In fact, they reveal a complex, temporal pattern wherein, for example, significant turbidite activity occurred at the end of MIS 6 (max. lowstand with sea-level of -100/-130m) as well as at the beginning of MIS 5 (max. highstand, sea-level of +7m) and yet almost no turbidites appear during the Last Glacial Maximum (MIS 2, sea-level of -120 m; Grant et al., 2014; Waelbroeck et al., 2002). Thus, no substantial maximum of turbidite sedimentation is detected during glacial low sea-level periods, as currently expected. Indeed, the significant turbidity current activity observed here duri ng i nte rglaci al sea-level highstand opposes the traditional sequence-stratigraphic interpretations that predict more turbidity current activity during sea-level fall and lowstand because of fluvial incision and increased fluvial transport to the marine basin (Posamentier and Vail, 1988; Kolla and Perlmutter, 1993, Piper et al., 1997). Our dataset suggests that changing sea-level conditions do not exert a primary control on the evolution of the sediment gravity deposits in the Zambezi deep-sea turbidite system. This is in agreement with the findings of other studies on highstand offshelf sediment transport (Boyd et al., 2008; Covault and Graham, 2010; Wiles et al., 2017a). It would be interesting for future research to investigate whether additional piston cores positioned higher up the continental sl ope (i .e. , cl oser to fluvial sources) show notable difference in turbiditic contributions between the glacial and i nte rgl acial periods or if they back up the findings of this study.
In order to test the possible impact of climate changes alone, in particular the effect of the southern African monsoon on the sediment delivery to the Mozambican margin, we reconstructed the terrigenous fluxes (MAR, see Section 3.5) at site MOZ4-CSF20, located northeast of the Zambezi river mouth in front of the central zone of the Mozambique continental shelf (18.5–16.5°S). Contrarily to sedimentation rates (SAR), the MAR allows to put aside the biogenic carbonate fraction that is related to primary productivity, and that significantly ‘pollutes’ the terrigenous contribution in low sedimentati on rate environments.. The MAR results reveal higher terrigenous inputs during interglacials and lower inputs during (full) glacials (see Section 5.2). This in line with the idea of dispersed northern inshore
transfer of terrigenous material on the Mozambican platform during interglacial periods, as recently proposed by Schulz et al. (2011), van der Lubbe et al. (2014) and Wiles et al. (2017a). Moreover, this orbital-scale pattern is also visible at higher resolution, during the MIS 5, with enhanced terrigenous inputs during the warm substages (MIS 5a, c and e) and associated sea-level highstands (Fig. 17D). Taken together with the timing for turbiditic deposition (i.e., irrespective of glacial or interglacial periods; se e Section 5.2), it confirms that sea-level conditions and the associated impact on the river-canyons connection do not solely control the continent-ocean sediment transfer on the Mozambican margi n . In contrast, our results suggest that warm conditions favored sediment transfer at this place. Hence, we assume that our results could reflect the direct influence of the southern African monsoon : e nhanced monsoon in SE Africa during interglacials (and low precession intervals, i.e., warm substages, in particular; e.g., Partridge et al., 1997; Simon et al., 2015; Johnson et al., 2016) might have caused episodic increases of the Zambezi water discharge and sediment load that in turn enhanced fine -grained fluvial sediment supplies towards the Mozambican margin. This interpretation is consistent with the monsoonal-related variability in sediment inputs observed on the Nile (Ducassou et al., 2009), Tanzanian (Liu et al., 2016), Congo (Picot et al., 2019) and Ganges-Brahmaputra (Weber et al., 1997) margins.
Interestingly, the occurrence of turbidites in the Zambezi turbidite system does not match the timing for enhanced terrigenous inputs off the Zambezi River (warm substages of the interglacial intervals ; Fig. 17). Thus, based on our piston cores we may speculate that the main initiation process for turbidity currents in the Mozambique Channel is the transformation (i.e., liquefaction) of failed sediment (e.g., slides, debris flow from retrogressive failures) from the continental slope. Turbidites of the Mozambique Channel could then be interpreted as slide-generated turbidites formed by typical fully turbul ent, l ow density, mixed sediment turbidity currents (see Piper & Normark, 2009 for a thorough review). Mass-movement features reported on the Mozambican outer shelf (Cattaneo et al., 2017) and the continental rise (Fierens et al., 2019) support this interpretation.
6.2.2. Distal to proximal migration of depocenter
Distally in the Zambezi depositional system, the cores from the Zambezi Valley and Zambe zi l ower fan are characterized by hemipelagic dark grey silty-clay that extends from MIS 13-14 (cores MOZ1-KSF30, MOZ2-KSF15 and MOZ4-CS25) to transition MIS 10/9 (i.e. Termination 4; cores MOZ2-KS11 and MOZ1-KSF30) (Fig. 18). This detrital facies, characterized by increased SAR (~3 cm/kyr, i.e., 3-fold increase in comparison to the younger deposits) at cores MOZ1-KSF30 (Fig. 9) and MOZ2-KSF15 (Fig. 11), encompasses both warm, humid (interglacials) and cold, arid (glacials) periods. At the same time, coarse, organic-rich turbidites are observed on the Zambezi Channel edge (core MOZ2-KS11), spi l l-over deposits are found on the channel levee (core MOZ1-KSF30), and coarse sands are deposited downstream from the channel outlet (core MOZ4-CS25) (Fig. 18). Thus, terrigenous sediment supply reached the distal parts of the Zambezi system between MIS 13-14 to the MIS 10/9 transition (i.e., Zambezi Fan in a “on” mode) (Fig. 19). On the contrary, the abrupt change in sedimentation to a yellowish beige sediment cover at all these distal sites after the MIS10/9 transition demonstrates the ceasing of terrigenous activity towards the distal depositional sites and, by extension, the distal Zambezi Fan (i.e. Zambezi Fan in a “off” mode) (Fig. 19). The age of 318±2 kyr at site MOZ2-KS11 (edge of Zambezi Valley) constitutes a minimum age estimate for this major sedimentation change.
Interestingly, an inverse pattern is observed in the Intermediate Basin. At site MOZ2-KS05, turbidite-starved yellowish foraminifer ooze (average SAR of 0.29 cm/kyr) dominated until 356±42 ka (Figs. 18 and 19) while fine-grained turbidites and the associated grey terrigenous facies (SAR of 1.96 cm/kyr) are deposited thereafter. In a more proximal location, core MOZ1-KS26 records a grey-terrigenous facies over the last ca. 500 ka (at least) (Figs. 18 and 19). Over this period, an increase in the turbidite frequency at site MOZ1-KS26 is observed at the same time that turbidites appeared at the distal site MOZ2-KS05 (Fig. 18). In other words, the increase of detrital inputs at MOZ1-KS26 is coeval with the
arrival of turbidity currents and detrital inputs at MOZ2-KS05. Taken together, these observations suggest a significant increase in the detrital supplies and the eastward progradation of the depositional system in the Intermediate Basin (Fig. 19).
Importantly, the age found for the increase of turbidity current activity in the Intermediate Basin (356±42 ka) corresponds within uncertainty to the age (> 318±2 ka) of the end of turbidity current activity, and of terrigenous supplies in general, in the deep system ~490 km southward (Figs. 18 and 19). Therefore, a major change in the feeding of the Zambezi depositional system occurred at ca. 356±42 ka. Before this date, terrigenous supplies occurred in the deep Mozambique Basin as well as in the (proximal) Intermediate Basin, while the Intermediate Basin becomes the unique depocenter for Zambezi-derived sediments after ca. 356±42 ka (Fig. 19).
6.2.3. Potential causes for the depocenter shift
Several possible forcings may be considered to explain the northward shift of the depocenter, from the distal Zambezi Fan to the Intermediate Basin, observed at about 356 ± 42 ka. Basically, this change requires the cessation of the feeding of the upper Zambezi Valley and, more generally, the southward rerouting of the Zambezi terrigenous flux to the Intermediate Basin. This suggests that the cause for the depocenter shift is located upstream of the upper Zambezi V alley, either at the continent-ocean transition (i.e., from the Zambezi alluvial plain to the continental shelf) or in the Zambezi catchment. Consequently, deep-sea forcings such as tectonic movements (e.g., through active faulting; Deville et al., 2018), sedimentological modifications (e.g., modification of sediment routing by a large mass transport deposit; Dennielou et al., 2019) or open-ocean oceanographic reorganizations (e.g., bottom-current reinforcement and the complete winnowing of the turbidite fl ows; Shanmugan and Moiola, 1982) are rejected, considering they cannot explain both the cessation of the turbidite activity in the Zambezi Fan and the increase in the terrigenous flux in the Intermediate Basin. The southward rerouting of the
Zambezi terrigenous flux to the Intermediate Basin at 356 ± 42 ka could, however, be explained by (1) a rerouting of Zambezi sediments on the continental shelf under changing coastal oceanographic conditions; (2) a significant change in the location of the Zambezi mouth through time or (3) a combination of the two.
Nowadays, and since the flooding of the continental shelf at ca. 11 ka (van der Lubbe et al., 2014), the Zambezi sediments are transported northwards along the Mozambique coast by large-scale cyclonic eddies. To the North, where the shelf is narrow, they finally reach the slope, close to the Zambezi Valley. The study of shallow cores along the Mozambican margin allows Schulz et al. (2011) and van de r Lubbe et al. (2014) to demonstrate that this pattern was significantly different before this time, with lowstand (glacial) conditions favoring the direct transport of the Zambezi sediments towards the Intermediate Basin. Hence, one could assume that the ‘post-flooding’ or highstand (interglacial) pattern for sedime nt dispersion on the Zambezi shelf dominated before 356 ± 42 ka, with a perennial feeding of the Zambe zi Valley and distal fan, while the lowstand pattern dominated thereafter. This could have been force d by the complex interactions between sediment inputs, delta progradation/retrogradation and oceanographic conditions, i.e., changes in the equilibrium response of the shelf system at long time scale. However, the northward transport of the Zambezi sediments to the Zambezi Valley (i.e. , highstand pattern; Schulz et al., 2011; van der Lubbe et al., 2014) does not seem to have an impact on the sedimentation in the Zambezi Valley and the Zambezi Fan, neither today (i.e., Holocene) or during previous interglacials (i.e., MIS 5, 7, 9, 11, 13, 15 and 17 in this study). This is shown by our reconstructions of the turbidite frequency in the Zambezi Fan that does not show any corre l ati on wi th sea-level changes (Fig. 18), and of the terrigenous flux at site MOZ4-CSF20 (Fig. 17). Indeed, the latter reveals that peaks in terrigenous flux over the last 150 kyr do not coincide with the highstand conditions (i.e., peak insolation maxima/minima in the northern/southern hemisphere, respectively) but wi th the maxima in local summer insolation ca. 11 kyr after (Fig. 17). Thus, the monsoon-related precipitation i s
invoked as the main forcing for riverine inputs (but not for the triggering of turbidity currents, at least at the studied sites) in the Mozambique Channel. If such, the climate evolution in Southern East Africa from a predominantly arid environment to generally wetter conditions over the last 1.3 Myr (Lyons et al., 2015; Johnson et al., 2016) could explain the increase of terrigenous inputs to the Intermediary Basin at 356 ± 42 ka. However, it cannot, alone, explain the attendant cessation of the feeding of the upper Zambezi Valley. Based on the above, we speculate that a southward migration of the Zambe zi de l ta i s needed to force the depcenter to shift northward, from the distal Zambezi Fan to the Intermediate Basin ca. 356 ± 42 ka.
The expected migration of the Zambezi mouth at ca. 356 ± 42 ka (or at least the exceeding of a latitudinal threshold if the migration was gradual) requires that the Zambezi delta was in a more northern position before this date. Considering the regional geological background, this implies that the lower Zambezi was running along the southern part of the Mozambique Belt, with a delta located ca. 100-150 km north of its current position, directly in front of the Upper Zambezi Valley. The current course of the Upper Zambezi Valley, initiated in the middle Miocene (Droz & Mougenot, 1987), could indicate that such a northern position of the Zambezi delta was dominant over the last 10-15 Myr, e ve n if multiple migrations of the delta should have occurred through time. The latter is likely supported by our data from the Zambezi Fan, and the alternating deposition of terrigenous-dominated muds (i.e., “on” mode of the Zambezi Fan) and biogenic-dominated muds (i.e., “off” mode of the Zambe zi Fan) at site MOZ2-KSF15 over the last 720 kyr. In details, this sedimentary record could suggest that the disconnection of the Zambezi delta from the Upper Valley known from ca. 356 ± 42 ka was also effective between ca. 644 ± 3 ka (cf. MOZ2-KSF15) and ca. 550 ± 20 ka (cf. MOZ4-CS25) (Figs. 18 and 19). In other words, we assume a northern position of the Zambezi delta before 644 ± 3 ka (and until 720 ka at l e ast) and between ca. 550 ± 20 ka (cf. MOZ4-CS25) and ca. 356 ± 42 ka (cf. MOZ2-KS05) (Fig. 19B), and a
southern position in between (i.e., from 644 ± 3 ka - MOZ2-KSF15- to 550 ± 20 ka - MOZ4-CS25-) and after ca. 356 ± 42 ka (cf. MOZ2-KS05) (Fig. 19B).
The forcing for the delta migrations may stem from complex origins, in which both autocyclic and allocyclic controls may play a role. Differential subsidence along the Zambezi margin could re pre sent a significant autocyclic control on the delta evolution, as previously shown for the Rhi ne (Cohe n, 2003), Nile (Stanley & Warne, 1993) or Yangtze (Chen & Stanley, 1995) rivers. The complex crustal structure of the study area, with a 23-km thick Beira High surrounded by a 7-km thick crust under the Lower Zambezi and continental shelf, and the Intermediate Basin (Mueller et al., 2016), could force a spatially and temporally non-linear rates of differential subsidence of the margin. This is corroborated by the complex evolution of the sedimentation and margin development off the Zambezi River throughout the Cenozoic (Walford et al., 2005; Ponte et al., 2018). At this time scale, the tectonic history of SE Africa (e.g., deformation of the East African Rift System; Chorowicz, 2005) and the attendant macroevolution of the Zambezi drainage network (Nugent, 1990; Moore & Larkin, 2001; Moore et al., 2012) have also h ad a profound effect on the delta evolution (Walford et al., 2005; Ponte et al., 2018). Thus, one could e xpe ct that the drainage evolution observed in the Zambezi catchment throughout the Pleistocene (disruption -reconnection of the Upper and Mid-Zambezi as a prominent example), although not precisely dated (i.e., Early to Mid-Pleistocene, Moore et al., 2012; Mid- to Late Pleistocene, Moore & Larkin, 2001), could force the delta migrations and the depocenter changes. To test this assumption, we tracked the s ource s of the marine sediments found in the Zambezi Fan and in the Intermediate Basin by usi ng ne odymium isotopic ratios that is a powerful tracer for geographical provenance (Goldstein and Jacobsen, 1988; Bayon et al., 2015). Interestingly, our results show remarkable homogeneity (Nd of -14.4 ± 1.8) over the last 720 kyr, both in the yellowish beige and dark grey sediments (Fig. 20A). Considering the high variability in Nd of the potential source areas in SE Africa (van der Lubbe et al., 2016; Roddaz et al., 2020), this suggests that the main forcing factor for the observed depocenter changes is certainly not
related to the evolution of the Zambezi catchment, or that large -scale Pleistocene changes in the Zambezi catchment (Moore & Larkin, 2001; Moore et al., 2012) pre-date 720 ka.
At this stage, it is not possible to constrain the precise forcing for the depocenter change s obse rved i n the offshore part of the Zambezi routing system. The acquisition of high-resolution seismic data on the continental shelf, and the recognition of paleo-channels on the Zambezi continental margin coul d he l p in deciphering the complex evolution of the sedimentation and the source -to-sink dynamics in the Mozambique Channel.
7. Conclusion
Deep-sea sediment cores from the Mozambique Channel allow to investigate the temporal distri bution and depositional patterns of terrigenous sediment in the Zambezi turbidite system over the last 720 kyr, and to identify the forcing mechanisms controlling the sediment inputs and their dispersal through time.
Our study reveal low sediment inputs and rare turbiditic deposits in the Zambezi turbidite syste m, and the deep (>2,500 m) Mozambique Channel in general, over the studied time interval. The occurrence of turbiditic deposits in the cores suggests that there is no genetic link between the temporal distribution, and hence triggering, of turbidites in the Mozambique Channel and evolving sea-level conditi ons ove r the studied period. According to the terrigenous flux reconstructed at high-resolution over the l ast 150 kyr in the upper part of the Zambezi depositional system, monsoon-related precipitation appears as the main forcing for riverine inputs in the Zambezi system. Yet, turbiditic deposits do not match the ti mi ng of increased riverine inputs. Thus, the transformation of failed sediment is proposed to be the main initiation process for turbidity currents. Finally, our results highlight regional-scale changes in sedimentary facies through time, interpreted as successive depocenter shifts within the Zambezi depositional system (and an on-off evolution of the distal Zambezi Fan) possibly rel ated to si gni ficant
changes in the location of the Zambezi delta through time. Our data suggest a northern location (compared to its modern location) of the Zambezi delta from ≥ 720 to 644 ± 3 ka and from 550 ± 20 to 356 ± 42 ka, associated with the delivery of detrital material toward the distal Zambezi Fan (in an “on” mode). The Zambezi delta could have migrated to a more southern location from 644 ± 3 to 550 ± 20 ka, and after 356 ± 42 ka (until today), causing a cessation of the feeding of the upper Zambezi Valley ( “off” mode of the Zambezi Fan) and a rerouting of the Zambezi terrigenous flux to the Intermediate Basin. The precise forcing for the delta migration is still not clear and would require further investigation.
This study highlights how complex and site-specific the study of such a great turbidite system is. It brings forward the great complexity of the depositional record in both temporal and spatial relationships of the Zambezi turbidite system, whose terrigenous inputs are controlled by climate and whose sediment distribution is the result of a complex and delicate balance between sediment transport/sedimentation processes and erosional/reworking mechanisms (themselves forced by a combination of autoge ni c and allogenic factors).
Acknowledgments
The PhD thesis of Ruth Fierens and Master thesis of Elisabeth Jorissen were co-funded by TOTAL and IFREMER as part of the PAMELA (Passive Margin Exploration Laboratories) scientific project. The PAMELA project is a scientific project led by Ifremer and TOTAL in collaboration with Université de Bretagne Occidentale, Université Rennes 1, Université Pierre and Marie Curie, CNRS and IFP-EN.
We thank the Captains and crew members of the 2014 and 2015 data acquisition during cruises PAMELA-MOZ1 and PAMELA-MOZ2 onboard the R/V L'Atalante and PAMELA-MOZ4 onboard of R/V
Pourquoi pas? and the technicians from Genavir that ensured acquisitions of the geophysical data. The
authors sincerely thank the Ifremer technicians (Angélique Roubi and Mickael Rovere) for their support
